The University of Miyazaki, NTT Data, and Pfizer have built an AI model that can be applied to electronic medical record data from multiple medical institutions to determine the effectiveness of drug treatment for lung cancer patients. This model performs natural language processing on unstructured data from electronic medical records to extract drug treatment effects.
In this research, we built a model using the large-scale language model BERT based on electronic medical record data from the University of Miyazaki, and then verified its applicability and practicality to electronic medical record data from six medical institutions. The results confirmed that the model can be applied to electronic medical record data from multiple medical institutions, and that clinical research evaluation items calculated from drug treatment effects extracted using this model show similar trends to results extracted by humans. did.
BERT is a natural language processing model announced by Google in October 2018.